(PeteCresswell) said:
Per G. Morgan:
I read the features write ups, but am too clueless to understand:
- If it supports redundant drives. i.e. Can it be set up so
that if a drive fails, data is not lost? Two drives?
That's what RAID6 is for.
http://en.wikipedia.org/wiki/Raid6#RAID_6
- If It supports hot-swapping. In the redundant-drive
scenario, can a bad drive be replaced on-the-fly,
or does the system have tb taken offline while the
array is rebuilt around the replacement drive?
Hot swap is a SATA feature. You check that the driver
type used, supports it. My only concern about the
concept, is handling a drive safely, so it doesn't
receive a physical shock when being swapped. And to
some extent, that's a physical design issue (packaging).
When a drive goes bad (on say, the RAID 6), the array
status is "degraded". You note carefully, which drive is
actually bad. Sometimes, you might have to stare at
the label on the drive, the serial number string, to be
sure. Next, you insert a replacement drive (hot swap).
The drive spins up. You enter the RAID management console.
Identify the new drive as being a new part of the RAID
set, and do a "rebuild". Parity blocks are computed for
the new drive, and the new drive then receives lots of
writes, while all the other drives are being read to do
the parity calculation. On a good RAID design, you can
"turn a knob", to limit rebuild bandwidth to a percentage
of array bandwidth. But even so (like at work once),
the results can be horrible. Our array needed to be
rebuilt in the middle of the day, and all home
directories were "slow" for the rest of the afternoon.
Teeth grittingly slow... 6MHz PC slow.
- Whether all the drives have tb the same size, or
1, 2, and 3-tb drives can be intermixed without losing
capacity on the larger drives.
- If there is a Hardware Compatibility List that specifies
which drives will work. Or will just about any old
SATA drive do?
You can use drives with TLER, to avoid nuisance declarations
of "degrade" or "fail". Other than that, a drive is a drive.
In terms of capacity mixing, that too is a function of the
driver. Intel had Matrix RAID, for supporting more
types on top of the same disks. Strictly speaking, if you
were doing something striped, the same amount of space would
be used from each disk. But if you can also run a second
array using the left-over space, there's no loss. Like
doing JBOD or something, with the leftovers.
See "Drive Extender" here, as another example of an implementation.
I think Microsoft may have removed "Drive Extender" from their
software now, so the functionality of this product has been gutted
(for the worse).
http://en.wikipedia.org/wiki/Windows_home_server
*******
When comparing FreeNAS to some packaged product you bought,
it's a matter of translating "features" of a commercial product,
to "implementation details" of a free product. The free product
can emulate the behaviors the commercial product has - it just
takes work, like configuration work, to make sure you've done
it all right.
And some things, just never work right on the free products.
I've *yet* to see a SAMBA/CIFS session on my Linux LiveCDs,
do anything at a decent transfer rate. So that's something
to watch for, and be aware of. For any product you're interested
in, you want benchmarks. There's no sense spending a month
tuning that crap, to find you're only getting 3MB/sec.
*******
NAS boxes can be as small as a single hard drive, in an enclosure
with a SOC chip. That's a "BYOD" bring your own disk box. But
those probably won't run very fast, and their main feature is
being a shared storage device. They don't necessarily have a
good backup story, or redundancy.
In terms of single drive types, there are "NAS" and "NDAS". NAS
uses well known standards (windows file sharing, NFS, FTP, HTTP,
or whatever). The NDAS on the other hand, the "D" stands for
"Direct". With an NDAS, each client computer needs a special
driver, as the protocol is non-standard. The main benefit,
is the protocol generally gives a higher transfer rate. But the
NDAS might not "play" with your TV, HTPC, iPad, the ice machine in
your refrigerator etc.
So there are some low end solutions, that are relatively cheap.
But if you're going to the trouble of setting up a server,
it might as well have a few features. Otherwise, a big-ass
USB flash stick might be just as good
(Sneakernet, walk
the files from one computer to another.)
I'm a big fan of Sneakernet, as I'm sick of wasting hours
trying to fix permissions problems, figure out why some
networking is busted, what the password is supposed to be
on a particular box, and so on. There's a lot to be said for
devices that you just plug in and use. Countless hours
can be saved.
Paul
